Reproduce “FunnelCausalNet: Funnel-aware Joint Conversion-Revenue Uplift for Multi-tier Coupon Allocation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yu Zhang, Zhihan Wang, Guanlin Chen, Min Jiang, Shuai Li

AMap Alibaba Group researchers developed FunnelCausalNet, a framework that jointly models user conversion and revenue uplift for multi-tier coupon allocation by explicitly encoding the deterministic funnel identity of Gross Merchandise Value (GMV). This approach reduced GMV effect-estimation error by 18-48% on semi-synthetic data and achieved the highest mean Delta ROI across incremental GMV targets on a large-scale industrial RCT dataset compared to other baselines.

Key insights

  • Explicitly modeling the funnel identity of GMV (product of conversion probability and conditional order value) is crucial for reducing variance and improving accuracy of heterogeneous treatment effect estimates, particularly under high zero-inflation.
  • Integrating funnel-aware uplift estimation with RCT-anchored Lagrangian relaxation provides a scalable and effective strategy for budgeted multi-tier coupon allocation, leading to improved ROI.
  • Joint split-conformal intervals, though conservative, offer a robust auditing layer for dual outcome uplift, enabling risk assessment and more responsible model deployment.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.